El « a priori» de la vía analítica: «una cosa es (…) en cierto modo

sessions predicting subsequent symptom change, controlling for symptom change up until that point. Table 11 gives the quadratic semi-partial rs each of these subscales at Sessions 4 and 7 predicting termination HRSD. I entered all individual predictors significant at the trend level (p < .10) into a regression. The starting model for Session 4 had five unique predictors (linear and quadratic PD and PE, and linear IPT) and an explanatory R2 of .57. After two iterations, the remaining model had an explanatory R2 of .50 and two predictors. The linear (semi-partial r [22] = -.46, p < .02) and quadratic (semi-partial r [22] = .41, p < .04) terms for PE remained significant, suggesting that moderate levels of process-experiential interventions were more predictive of outcome than were more higher or lower levels of process-experiential interventions. For Session 7, the starting model had an explanatory R2 of .72 and all four unique predictors in the model were statistically significant. Moderate levels of psychodynamic (for the linear term, semi-partial r [20] = -.42, p < .05; for the quadratic term, semi-partial r [20] = .42, p

< .05) and process-experiential (for the linear term, semi-partial r [20] = -.52, p < .01; for the quadratic term, semi-partial r [20] = .53, p < .01) predicted better outcome than did higher or lower levels of these interventions.

Discussion

Therapy process description. As with Study 1, some of the complexities that

exist in the process and outcome of a different type of therapy, psychodynamic, and for a different disorder, depression, were uncovered. Psychodynamic interventions were prominent in psychodynamic therapy but their levels were relatively low across two early sessions as assessed by independent observers (i.e., rated as only “somewhat typical” of

the sessions). What these levels of dynamic therapy interventions might mean for the concept of psychodynamic adherence cannot be easily ascertained from this one study. These ratings may be lower than those provided by direct participants in the therapy process (i.e., clients and therapists). Perhaps observers lack the emotional cues available to direct participants in the therapy process that might amplify or make the

psychodynamic interventions more salient.

Individual psychodynamic techniques were also not employed uniformly. Transference interpretations, the tools the dynamic therapist uses to putatively bring about change, were used quite sparingly (for reviews, see Crits-Christoph & Gibbons, 2001; Hoglend, 2004). Most of the work in dynamic therapy appeared instead to be collecting information about the client’s experiences and interpersonal patterns. The lower levels of transference interpretations relative to exploratory interventions might be for several reasons. First, while early interpretations are encouraged by some dynamic theorists (Gill, 1982; Malan, 2001; Schafer, 1983), actual opportunities to make

interpretations might be limited. Dynamic therapists may want to collect a large database about the client’s interpersonal and intrapsychic life before offering an interpretation so their intervention is accurate. The accuracy of transference interpretations to the client’s CCRT has been shown to be related to better therapeutic outcomes (e.g., Crits-Christoph, Cooper, & Luborsky, 1988). Alternatively, clinical timing may play an important part in the therapist’s decision to intervene with an intervention. The therapist might very carefully choose when to give the interpretation based on the client’s level of defensiveness.

A second reason for the judicious use of interpretations might be due to therapists’ awareness of their “high cost, high yield” association with outcome, either from clinical experience or based on the research literature (e.g., Crits-Christoph & Gibbons, 2001; Hoglend, 2004). Transference interpretations may have the potential to produce significant positive change in the client but may also be disruptive when

mishandled. Too many interpretations, even if they are accurate, may be too destabilizing to the client. Dynamic therapists may choose to wait before delivering an interpretation if they recently used another interpretation in the session. Further study of the process of psychodynamic therapy might reveal patterns in the process of therapy that determine the use of interpretative versus explorative techniques.

Common factor interventions were the most highly endorsed of all the

intervention types in this sample and obviously play a significant role in psychodynamic therapy, especially in supportive-expressive psychotherapy. Even so, they were still lower than might be expected (rated not quite “typical” of the session on average). Perhaps the lower-than-expected levels of common factor interventions might be due to how the supportive component of supportive-expressive psychotherapy is defined. Dynamic therapists do provide empathy and positive regard for their clients, but they might more specifically work to support ego functioning, or the ways their clients protect themselves from unacceptable wishes or negative feedback (e.g., rationalization of poor performance by suggesting one did not like that activity). These interventions may not be reflected in the definitions proposed by many common factors advocates (Frank & Frank,

1991; Wampold, 2001) and so might not be captured in the MULTI items making up the common factor interventions subscale.

The therapists in this study frequently borrowed interventions from person- centered and process-experiential therapies that explored clients’ experiences and the meaning clients ascribed to them. These interventions might indicate that the exploration in psychodynamic therapy is not limited to relationship and interpersonal factors, but may include more of the client’s entire phenomenology, including emotions, sensations, and personal inclinations. There is little doubt that many modern dynamic therapists believe that deepening affect is very important (Blagys & Hilsenroth, 2000; Fosha, 2002; McCullough et al., 2003; Summers & Barber 2010; Wachtel, 1993), and perhaps what I observed represents the merging of different techniques to further this goal in dynamic therapy. Bringing person-centered and process-experiential theory to bear on the practice of dynamic therapy might provide a more comprehensive understanding of the

exploratory interventions in dynamic therapy and might open up new directions for process research (cf. Diener, Hilsenroth, & Weinberger, 2007).

Psychodynamic therapists avoided interventions that directly instructed or encouraged clients to change their behavior. Most cognitive-behavioral interventions were therefore relatively low in intensity and frequency. Those cognitive-behavioral interventions that were used tended to be supportive in nature (e.g., identifying coping strategies). Directive interventions such as provided in interpersonal psychotherapy, like encouraging the client to join a group in order to make new friends or to consider others’ wants before acting, were also similarly eschewed. This trend most likely suggests that

therapists in dynamic therapy do not attempt to change behavior directly but rather

choose to help develop their clients’ insight into their interpersonal patterns with the hope that clients will begin to change their behaviors with this knowledge.

Correlations between therapy process and outcome.

Correlations of specific factors to outcome. Techniques specific to the psychodynamic treatment manual were related to symptom change in this study of dynamic therapy for depression. However, this association was not simply that greater levels of psychodynamic interventions predicted greater subsequent outcome. Moderate levels of psychodynamic interventions at both Sessions 4 and 7 were more related to subsequent symptom improvement than were higher or lower levels of psychodynamic interventions. A curvilinear relation had been hypothesized (Ogrodinizuk & Piper, 1999; Barber et al., 2008) and observed (Barber et al., 2008 [albeit in the opposite direction]) in previous investigations of dynamic therapy for substance dependence. The amount of psychodynamic interventions used with a particular client might in some way bring about the client’s eventual outcome. Moderate levels of psychodynamic interventions might represent a precise dose of interpretative comments that help the client organize his or her experiences and begin to change his or her interpersonal patterns based on this new understanding. Lower levels of dynamic interventions may not be enough to trigger the development of insight, either because the therapist did not engage the client in thinking about his or her interpersonal and intrapsychic life or did not provide the connections between past experiences and present feelings and behaviors. High levels of dynamic interventions might confuse the client with too many directions for exploration or might

be toxic in high doses by creating too much anxiety or shame for the client, disrupting his or her way of managing painful or unacceptable experiences.

The level of psychodynamic interventions might alternatively indicate some qualities of the client or the therapy process that are related to outcome. Very low levels of psychodynamic interventions could be due to poor fit between the client and the treatment approach (e.g., low ability to introspect, unwillingness to tolerate therapeutic silences and ambiguity, demanding solutions from the therapist). On the other hand, very high levels of dynamic interventions might indicate that the patient is interpersonally difficult and not responding to treatment. Alternatively it may reflect a situation in which a therapist feels pressured to act by the client, perhaps repeating the client’s problematic interpersonal patterns rather than providing an opportunity to change them (e.g., a client with a tendency to intellectualize relishing the discussion of his or her interpersonal life, a dependent client who believes accepting whatever the therapist says will sustain the relationship, a passive-aggressive client who rejects the therapist’s interventions but accuses the therapist of not being helpful). Moderate levels of dynamic techniques might then be used with more easily manageable clients without such traits or who are likely to improve, perhaps even regardless of the interventions used.

Correlations of common factors to outcome. Surprisingly, none of the common

factors that measured in this study (neither interventions nor alliance) were related to outcome (see also Barber et al., 1996). This finding is especially surprising because supportive interventions and the therapeutic relationship are given a high place in psychodynamic therapy (e.g., Greenson, 1967; Luborsky, 1984). The common factors

assessed in this study might not have been the same as the supportive interventions provided by the dynamic therapists or might have been obscured by the therapists’ commitment to therapeutic neutrality. Intervention measures that include more ego- supportive techniques in addition to general relationship enhancement techniques might be better able to reveal any relation with outcome, although intervention scales specific to supportive interventions in dynamic psychotherapy have failed to find any such relation (e.g., Barber et al., 1996; Barber et al., 2008).

Therapeutic alliance has been shown to be related to outcome in other studies of dynamic therapy (for a review, see Crits-Christoph & Gibbons, 2003; for exceptions, see Barber, 2009). In this study, however, the effect of alliance on subsequent symptom change might have been overshadowed by prior symptom change (Barber, 2009). For instance, at Session 7, prior symptom change was significantly related to alliance scores, and without controlling for early symptom change the relation of Session 7 alliance and outcome is marginally significant. This statistical relation may make it difficult to detect any relation between alliance and subsequent outcome. Such relations between early symptom change and alliance might also indicate that good alliances are a result of clients feeling better in therapy in this sample. As a product of prior symptom change, alliances might not be as predictive of subsequent outcome as what originated them.

I was not positioned to test for a mediating effect of alliance on the relation of common factor interventions on outcome because in this sample common factors (either common factor interventions or alliance) were not related to subsequent outcome. Other future studies will need to investigate this hypothesis.

Correlations of unintended factors to outcome. Moderate levels of process-

experiential interventions predicted better outcome at both Sessions 4 and 7 than did lower or higher levels of process-experiential interventions. This curvilinear relation impacted outcome more than the contribution of other types of interventions, including psychodynamic interventions at Session 4. The most common process-experiential interventions used by the dynamic therapists in this study were ones that explored the client’s affect and ambivalence. These interventions may work directly on alleviating clients’ symptoms through the expansion and creation of new affects in the therapy setting. They have also been considered a feature of good dynamic therapy as feelings may be the landmarks therapists use to successfully assess or intervene on clients’ motivations and interpersonal conflicts (Blagys & Hilsenroth, 2000; Malan, 2001; McWilliams, 1999).

Moderate levels of process-experiential interventions may allow the client to access painful or repetitive feelings and may provide guidance toward new ways of generating more adaptive emotional experiences (e.g., accepting the anger the client feels is unacceptable toward a lost loved one). Low amounts of process-experiential

interventions may not activate the client sufficiently for the material to be meaningful to the client. They may also call the client’s attention to his or her emotions but not provide any additional guidance for the client to create a new emotional experience, leaving the client stuck in his or her problematic feelings. Low levels of process-experiential interventions might also be used with clients who are alexithymic or avoidant of

levels of process-experiential intervention may evoke very strong emotions in the client and be disruptive to his or her ability to regulate emotions or reflect on the information they provide. High levels of process-experiential techniques might also lead clients to view their therapists as stereotypical and to reject their interventions, especially male clients who might feel uncomfortable with an emotion-focused approach.

General Discussion

These two studies used descriptive and correlational approaches to reveal some of the complexities in the process of two different psychotherapies for two different

disorders. These complexities, like variation in the levels of interventions from the treatment manual and the use of techniques outside the treatment manual, occurred even though both therapies investigated were manualized and monitored. The studies also showed that these complexities in psychotherapy process affected subsequent client outcome. Treatment outcome comparisons, invaluable for their documentation of the efficacy of psychotherapy, have not directly tested the relation of varying levels of specific and common factors to outcome and have often overlooked the contribution of unintended factors toward symptom change. They are therefore limited in what can be inferred from them about the mechanisms of change. Descriptive and correlational

process research might then be better able to inform us how psychotherapy works, both in terms of what specific, common, and unintended factors are in therapy and how they are related to outcome.

Specific Factors

Specific factors, or those interventions consistent with the treatment manual employed, were associated with better outcome in both behavioral and psychodynamic therapy. However, it was not simply that greater levels of specific factors were related to greater symptom improvement. Rather, moderate levels of the specific factors were related to better subsequent outcome compared to higher or lower levels. Most treatment outcome comparisons have not directly examined the relation between differing levels of

technique use and outcome. However, what constitutes an appropriate level of specific technique use for a study to examine is not known from theory or empirical studies, and it is possible that some treatment outcome comparisons are not testing the optimal version of their therapy (i.e., one with a moderate level of specific technique use). Again, there is little theoretical or empirical work that would help choose the levels of interventions for these conditions. Further descriptive analyses of psychotherapy process can establish the range of specific techniques that therapists use in a treatment, and correlational studies can test how outcome is influenced across a full range of technique use.

The descriptive components of these studies have shown that individual specific technique use varies in behavioral and psychodynamic treatment, even if it is also true that interventions from the same theoretical orientation tend to co-occur in the session (i.e., exhibit high Cronbach α coefficients). The primary interventions essential to the theory behind the treatment were not always the most frequently used or typical interventions in practice, and the use of auxiliary interventions ranged widely. For instance, in behavioral therapy, exposure (item 16) was scored highly by clients and therapists, but homework assignment and review (item 17) was even more frequently rated. Skill-building (item 15) was considered to be relatively less representative of the sessions as were those other techniques. In psychodynamic therapy, interpretation (items 2 and 22) was less often used than many other interventions. Exploratory interventions (items 14, 19, and 40) were in fact much more prevalent than these core techniques. Relying then on only a handful of techniques to define adherence does not recognize the

complexity of psychotherapy process, and more descriptive work on which individual interventions compose a treatment is needed.

Even though curvilinear relations between specific factors and symptom change were found for two very different samples in this thesis, it remains unclear how these curvilinear functions might generalize to other samples and contexts. One possibility is that these functions represent absolute relations between scores on the specific MULTI subscales and subsequent outcome. In such case, the same curvilinear function would be expected to generalize across all samples of that particular theoretical orientation. When given a MULTI subscale score for any particular session of that orientation, we would be able to classify the session as having low, moderate, or high adherence and predict the approximate amount of symptom decline the client would experience by the end of therapy. Differences in the range of intervention use in a sample might determine which section of the curvilinear relation with outcome is observed. For example, if the relation between client-rated BT subscale scores and outcome from Study 1 were generalizable across all samples of behavioral therapy for OCD, a sample in which BT subscale scores ranged only from “typical” to “very typical” would exhibit a positive correlation with symptom improvement, whereas a sample with a range from “not at all” to “slightly” would most likely evidence a negative correlation with outcome. Absolute

generalizability across samples would provide an easily interpretable meaning for

adherence scores based on their expected relation with outcome. It would also provide an explanation for the mixed findings of previous correlational studies of adherence and outcome. However, absolute generalizability of any curvilinear findings is unlikely given

the complexity of the process of psychotherapy and the wide variability in how the same therapy is practiced.

Another possibility is that curvilinear relations between adherence and outcome are relative to site or cohort. Contact among therapy providers or raters might (explicitly or implicitly) lead to a valuing of moderate levels of interventions over more extreme levels of interventions. Therapists at the same clinic, graduating from the same training program, or receiving the same supervision may develop a consensus as to what

constitutes good therapy, and this consensus may be for moderate intervention use. Similarly, raters undergoing reliability training may settle on an agreement as to what levels of adherence are moderate and what levels are more extreme, perhaps based on the outcomes they imagine to occur from the sessions they are rating. The consensus reached may vary from by site or cohort, but moderate levels of adherence, however they were defined within that group, may always be associated with greater improvement than more